Especies de bambúes a utilizar y propiedades –Uso del bambú local

The classes used the Elluminate® CSCL Tool, which has a Whiteboard, IM, Application Sharing, Web Access, VoIP and integrated video streaming of lectures and WebCT’s Bulletin Board. The students were a mixture of on-campus and distance students (full time students and working professionals) and there were 358 students in all the eight courses. These courses were GMPM – Global Marketing and Product Management, CDW – Communication Design for WWW, FHCI – Foundations of HCI Usability, PP –Proposing and Persuading, ETC – IT and Decision Systems Capstone, SD – Studio Design in HCI, IB – International Business and TCTR – Theory and Research in Technical Communication. Of the various network measures that can be used to understand how information flows between nodes in a network, the most common measures are centrality measures which measure the importance of the node in the network and of these degree centrality, closeness centrality, and betweenness centrality measures are often cited in literature. Additional centrality measures are reach centrality, eigenvector, power, information and influence centralities. Other network measures such as cohesion (presence of subgroups/cliques), structural holes, network density, network transitivity etc. are used to understand individual and group behaviour in networks of different sizes. We will restrict our discussion to centrality measures as they are sufficient to illustrate the emergence of new actors in the network taking on leadership roles in directing class discussions and knowledge sharing.

As the name indicates, degree centrality indicates the number of lines coming in and out of a node (number of ties – in degree and out degree). When looking at friendship or knowledge sharing situations, we can look at in degree as an indication of popularity and out degree as an indication of extreme social behaviour willing to get in touch with many network alters (sending out friendship requests on Facebook®). Closeness centrality (based on the mean geodesic distance – shortest path between two nodes and can propagate via other network nodes), is useful to measure nearness of two nodes in the topological space where distance measurement is rather difficult. This shortest path is relative to other short paths in the network between these nodes in question. Betweenness Centrality, while also based on the shortest paths between nodes in a network identifies network nodes (actors) that occur on many shortest paths between other nodes. The interpretation is that network nodes (actors) with high betweenness centrality scores can act as ‘controls’ or ‘conduits’ for information flow between nodes in the network. For this reason I have chosen to focus on the betweenness centrality scores to indentify actors that as acting as ‘conduits’ for information flow in the network and hence facilitate knowledge sharing. Since Eigenvector centrality scores also appear in the tables below, it is important to know that eigenvector centralities indicate the importance of the node in the network, for example consider the PageRank® feature in Google searches.

The following figures and tables show the sociograms for some of the chat and bulletin board conversations and the centrality scores for the actors respectively (the betweenness centrality scores have been highlighted). UCINET software (Borgatti, Everett & Freeman 2002) was used calculate the centrality scores and map the network diagrams. While we look at individual actor centrality scores to get an indication of an actor’s position in the network based on tie direction, nearness, betweenness etc., how these centralities are related provides more information about the whole network. For this we look at how centralized a network is (one that has one or a few actors very central to the network) or how decentralized a network is (many central actors). Because a highly central node can be a single point of failure in a network, fully centralized nodes with one or a few well connected hubs (those nodes with high betweenness and degree centrality scores) can fail suddenly. Networks that are more decentralized (sparse networks) have many central actors and do not have single points of failure. These networks provide multiple pathways via central actors for information flow, can withstand attacks to the network, and do not tend to fail suddenly. Chat and bulletin board networks can take on either property (highly centralized or highly decentralized), depending on the context. The general observed behaviour of people in chats and bulletin boards is that only a few participate regularly and with high frequency leading to fairly well centralized networks. However, since many of these networks are created ad hoc (for a project, a class etc.) they need not be sustained for long periods of time, so failure may not be an issue. However, if the prominent actors in a chat room stop writing or posting, then it is possible that the network may cease to exist after a while or newer actors will step up to take on the roles of those who have left the network. With this perspective, we can start looking at the chat and bulletin board networks for these classes/courses. While figure 1 indicates four

Binod Sundararajan

central actors (drgj-the instructor, ksks, calg and grj), the betweenness centrality scores suggest that apart from the instructor drgj, it is ksks who has a higher betweenness centrality score than his peers. However, the sociogram reflects only a small percentage of students enrolled in the class. Many did not participate actively in the chat discussions substantially, other than occasional queries.

Figure 1: GMPM - Global Marketing and Product Management (chat sociogram)

It is interesting to note that the actor with the next highest Betweenness Centrality score, grj (26.667), while central, was more central in the bulletin board post network. She was the most prolific poster with over 52 of the 115 or so posts during the term. This actor was ready to help others in the network by providing answers to questions posed. However when asked who in the network she learned from and found knowledgeable, she mentioned that she did not learn from anyone else in the class. So it is interesting to see her centrality scores in the chat sociograms shown in figure 2, Degree Cent (40), Closeness Cent (52.632) and Betweenness Cent (26.667) and Eigenvector Centrality (64.683). Compare that with the scores of actor Ksk (Deg – 40, Closeness – 62.5, Betweenness – 46.667 & Eigenvector – 69.545). Viewing the bulletin board post sociogram will put this in perspective and this can be seen in figure 2. Here actor grj (denoted as gren) is most definitely the central actor, with actor ksk appearing outside the network in conversation with actor calg (denoted in Table 2 and Fig 1 as clg).

Figure 2: GMPM - Global Marketing and Product Management (bulletin board sociogram)

Table 2: GMPM - Global Marketing and Product Management (centrality scores)

The reading here can be that each actor might prefer one medium over the other. The bulletin board medium is asynchronous, giving actors time to compose their replies, while chat conversations are synchronous requiring actors to be able to respond to queries quicker. While some may prefer one medium (depending on the cue richness of the medium), over another, both of these allow participants to express their ideas and viewpoints and get them engaged and involved with the material or the topic of the discussion.

Turning our attention to the other classes (courses) we see a variety of centralized and decentralized networks. The PP class (a smaller class) on the other hand, shows a fully connected network and there was no clear central actor.

Figure 3: PP – Proposing and Persuading chat sociogram class 1

The fully connected network has the property of almost all the actors knowing exactly what the rest of the actors in the network know. It is said that no new information is created or propagated in such networks, while sparse networks are considered best for information propagation (friend of a friend of a friend – FOAF). Examples are trust and family networks vs. FOAF networks (LinkedIn® is one such example). This can be seen in the centrality scores shown in table 3.

Binod Sundararajan

Figure 4: PP – Proposing and Persuading chat Sociogram class 9 Table 3: PP – Proposing and Persuading – centrality scores

As can be seen from the centrality scores in Table 3, and the fully connected network sociograms in figures 3 and 4, there are more than a few central actors. This being a small network (class size was around 12-15, with the remaining actors not participating in any discussions), the few actors that are central (anybody who posts anything can become central) share ties, closeness and importance. However, we do note that there are a few who occupy ‘conduit’ positions indicating information exchange through these highly between nodes in the network.

Figures 5, 6 and 7 depict the chat sociograms for the CDW course for classes 1, 3 and the last class respectively. In addition to the “chair” (the TA or the instructor), actors Chris, Will and Teresa occupy central roles in the class chat discussions. This can be seen by their high betweenness centrality scores in tables 4, 5 and 6. What is interesting, however, is as the semester progressed, while actors Chris, Will and Teresa continue to hold fairly central positions in the chat discussion networks, other actors emerged throughout the semester to occupy fairly equal roles.

Figure 5: CDW – Communication Design for WWW – chat sociogram class 1

This can be seen by the emergence of actors Brian, Jason and Ian (Figure 5 and Table 5) and actor Dale (Figure 6, Table 6). Consequently the betweenness centrality scores of actors Chris, Will and Teresa appear to decrease as the semester progressed indicating that other actors were allowed to not only voice opinions but also take up central roles in the discussions.

Binod Sundararajan

Figure 6: CDW – Communication Design for WWW – chat sociogram class 3

The “chair” actor (instructor or TA) did lead several discussions and answer queries, content analysis of the conversations indicate that student actors did have a considerable say in how the discussions continued. This was evident in conversations in almost all the courses, not just the CDW course. Table 5: CDW – Communication Design for WWW – centrality scores class 3

Figure 7: CDW – Communication Design for WWW – chat sociogram last class

Continuing to look at other classes, we note the same phenomenon repeating itself as during the course of the semester new actors emerged to take on central roles in class discussions.

Table 6: CDW – Communication Design for WWW – centrality scores last class

Interestingly, the bulletin board discussions for the CDW course were almost non-existent, indicating that possibly this set of students preferred instant messenger interactions to the asynchronous bulletin

Binod Sundararajan

board discussions or maybe the course material may not have warranted much asynchronous discussions.

Figure 8: FHCI – Foundations of HCI – chat sociogram

Figure 9 depicts the rather sparse sociogram of the FHCI course bulletin board discussions, indicative of the general trend of bulletin board discussions. For the sake of space, only some of the sociograms of the various courses are presented. This is contrary to the behaviour seen in the chat sociogram for the same class (figure 8). There is a lot more activity in the chat conversation. This is to be expected as students are sitting in for class – co-located or from a distance and would like many doubts clarified as they keep pace with the lecture.

Figure 9: FHCI – Foundations of HCI – bulletin board sociogram

Table 7: FHCI – Foundations of HCI – centrality scores

Of all the bulletin board discussion sociograms in the eight courses, ETC was the only course that had a significant number of discussion posts and figure 10 depicts the most active of the discussion threads. We note that chat conversations during the lecture are not only productive, but preferred. Discussion board or bulletin board conversations are generally sparse and unless specifically driven (as in the case of the ETC course), tend to have at the most 1-3 posts on average and very little by way of sustained discussion. Network position (betweenness centrality) has a significant relationship with influence obtained among peers.

Binod Sundararajan

Table 8: ETC – IT and decision systems capstone – centrality scores

Figure 11: ETC – IT and decision systems capstone – bulletin board sociogram